Annual Reviews Solid

ANRV340-PC59-11 ARI 26 February 2008 21:0
Advances in Correlated
Electronic Structure
Methods for Solids,
Surfaces, and
Nanostructures
Patrick Huang
and Emily A. Carter

Department of Mechanical and Aerospace Engineering and Program in Applied
and Computational Mathematics, Princeton University, Princeton,
New Jersey 08544-5263; email: eac@princeton.edu
Annu. Rev. Phys. Chem. 2008. 59:26190
First published online as a Review in Advance on
November 19, 2007
The Annual Review of Physical Chemistry is online at
http://physchem.annualreviews.org
This articles doi:
10.1146/annurev.physchem.59.032607.093528
Copyright c 2008 by Annual Reviews.
All rights reserved
0066-426X/08/0505-0261$20.00
Current afliation: Lawrence Livermore

National Laboratory, Livermore, California 94551
Key Words
density functional theory, excited states, electron correlation,
embedded cluster models
Abstract
Calculations of the electronic structure of solids began decades ago,
but only recently have solid-state quantumtechniques become suf-
ciently reliable that their application is nearly as routine as quantum
chemistry is for molecules. We aimto introduce chemists to the pros
and cons of rst-principles methods that can provide atomic-scale
insight into the properties and chemistry of bulk materials, inter-
faces, and nanostructures. The techniques we review include the
ubiquitous density functional theory (DFT), which is often suf-
cient, especially for metals; extensions such as DFT+U and hybrid
DFT, which incorporate exact exchange to rid DFT of its spuri-
ous self-interactions (critical for some semiconductors and strongly
correlated materials); many-body Greens function (GW and Bethe-
Salpeter) methods for excited states; quantum Monte Carlo, in prin-
ciple an exact theory but for which forces (hence structure optimiza-
tion and dynamics) are problematic; and embedding theories that
locally rene the quantum treatment to improve accuracy.
261
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
DFT: density functional
theory
TDDFT: time-dependent
density functional theory
QMC: quantum Monte
Carlo
1. INTRODUCTION
We are interested in a predictive theory of solids. Although no single theory exists
that gives all observables for every material type, the combination of efforts from
various disciplines has resulted in a portfolio of methods, each with their respec-
tive strengths and weaknesses. In this review, we attempt to summarize some of the
currently available methodologies to characterize the electronic structure of solids
and related phenomena (e.g., chemisorption). We emphasize the essential approxi-
mations that enter into the different approaches and point out their successes and
failures.
The discussion begins with density functional theory (DFT) (13), which is by
far the most widely used tool for the ground-state properties of solids, surfaces, and
nanostructures. Much of this popularity is because DFT provides a good balance
between simplicity and accuracy. Despite its successes, there are many well-known
situations in which DFT is unsuitable, and in this review, we present alternative
methods that address some of these deciencies.
Although information fromDFTis often interpreted as approximations to excited
states, DFT is formally a ground-state theory. We discuss several routes to excited
states that are more rmly grounded in theory. The rst is the GW method (46)
for quasiparticle excitations, which are transitions between N- and (N 1)-electron
states. These directly correspond to excitations probed in photoemission (electron
removal) and inverse photoemission (electron addition) experiments. The second
class of excited-state methods includes time-dependent DFT (TDDFT) (7) and the
Bethe-Salpeter theory (6), which yield neutral particle-hole transitions associated
with optical experiments.
DFT also performs poorly in situations involving strong electron correlation ef-
fects. The quantumMonte Carlo (QMC) method (810) caninclude electroncorrela-
tion for periodic systems and can also access individual excited states. In the solid-state
community, QMC is largely regarded as essentially exact, although the calculations
are rather demanding. As a result, we also comment on less-expensive approaches for
strongly correlated electrons, including extensions to DFTsuch as hybrid functionals
and DFT + U schemes.
Lastly, we discuss embedded cluster models for the solid state, which have a long
history in the study of chemisorption. The philosophy behind these models is to sub-
divide the system into a region of interest described at higher level of theory, whereas
the surroundings are accounted for in a more approximate manner. A partitioning
of the model into different regions requires a choice for the inter-region coupling.
We summarize the most common models, including, e.g., purely electrostatic and
DFT-based embedding potentials.
Of course, owing to space limitations, we cannot discuss many other promising
avenues of development. These include the dynamical mean eld theory to deal with
strongly correlated materials (11), recent extensions of traditional quantumchemistry
approaches to periodic systems (1215), and advances in O(N) electronic structure
methods (16), just to name a few.
262 Huang
Carter
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
HF: Hartree-Fock
XC: exchange correlation
2. DENSITY FUNCTIONAL THEORY
The early foundations of DFT (13) are due to Hohenberg and Kohn (17), who
established the use of the density (r) as the basic variable from which all ground
state properties, in principle, can be derived. In particular, they showed that there
exists a universal energy functional E[], whose minimum corresponds to the exact
ground-state energy E
0
. Although this represents an enormous simplication of the
interacting many-electron problem, unfortunately, the exact functional dependence
of E[] is not known.
Kohn & Sham (18) subsequently devised a practical scheme by mapping the prob-
lem of N interacting electrons in an external potential
ion
onto a ctitious set of
N noninteracting electrons in an effective potential
s
. Here, we regard
ion
as the
electron-ionpotential providedby the nuclei, althoughthe theory holds for more gen-
eral external potentials. The noninteracting problem is well-known from Hartree-
Fock (HF) theory: The total energy is an explicit functional of the one-electron
orbitals
j
, and has the form
E[{
j
}] = T
s
[{
j
}] +J[] + E
xc
[] +
_
dr(r)
ion
(r), (1)
where (r) =

occ.
j
|
j
(r)|
2
is the density, T
s
is the exact, noninteracting kinetic
energy,
T
s
[{
j
}] =
1
2
occ.
j
|
2
|
j
, (2)
and J is the classical Hartree energy,
J[] =
1
2
_
dr dr
(r)(r
)
|r r
|
. (3)
E
xc
is referred to as the exchange-correlation (XC) energy and is dened as containing
all remaining quantum effects not captured by J and T
s
. This quantity is generally
unknown; however, useful approximations for E
xc
do exist. The use of an approximate
E
xc
represents a major source of uncontrolled errors, for there is no general strategy to
systematically improve on E
xc
. The search for high-quality approximations for E
xc
is a
subject of ongoing research. Additionally, an active area of research is to forgo orbitals
entirely and employ instead an approximate T
s
[] that depends explicitly on , and
employ approximations to both T
s
[] and E
xc
[]. This forms the basis for an orbital-
free DFT scheme, which avoids the computational expense of dealing with orbitals
(19, 20). Unfortunately, no generally applicable T
s
[] is available; hence orbital-free
DFT is still limited in scope.
Minimization of Equation 1 with respect to the orbitals
j
leads to
_
1
2
2
+
s
[](r)
_

j
(r) =
j
j
(r), (4)
www.annualreviews.org Solid-State Electronic Structure Methods 263
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
KS: Kohn-Sham
LDA: local density
approximation
GGA: generalized gradient
approximation
where
j
are the orbital eigenvalues, and
s
is the effective Kohn-Sham (KS)
potential,
s
[](r) =
J[]
(r)
+
E
xc
[]
(r)
+
ion
(r). (5)
The eigenvalue of the highest-occupied KS orbital can be identied with an ioniza-
tion potential or work function. But beyond this, KS wave functions and eigenvalues
are not guaranteed to yield actual one-particle quantities for the real, interacting sys-
tem, and interpretations based on the explicit use of KS orbitals should be made with
caution. It is common to report the DFT energy gap
g
=
N+1
N
, where
N+1
and
N
are the eigenvalues of the lowest-unoccupied and highest-occupied states obtained
from a KS calculation for the N-electron ground state, respectively. Although Koop-
mans theorem provides a physical interpretation for HF eigenvalues as one-electron
removal/addition energies, the KS eigenvalues do not correspond to excitation ener-
gies of any sort. It can be shown (2123) that
g
differs from the experimental band
gap by the derivative discontinuity , which represents the discontinuity of the XC
potential with respect to a change in particle number. The presence of does not
result from a failure of approximate XC functionals, and it remains nite even if the
exact XC is known. Little is known about the magnitude of in a realistic solid,
although it has been suggested that it could constitute a signicant portion of the ex-
perimental band gap (24). In Sections 3 and 4, we provide an overview of methods to
treat excited states in solids that have a rmer theoretical basis than an interpretation
of KS eigenvalues.
2.1. Local and Semilocal Approximations
By far the most common class of approximate XCemployed in solid-state problems is
the class based on the local density approximation (LDA) and the generalized gradient
approximation (GGA). In the LDA, the XC energy is a local functional of the density
and has the form
E
LDA
xc
[
] =
_
dr(r)
unif.
xc
[(r), (r)], (r) =

(r)
(r)
(r)
, (6)
where represents the spin polarization density, and
unif.
xc
=
unif.
x
+
unif.
c
is the
uniform electron gas correlation energy per particle, consisting of an exchange
unif.
x
and correlation
unif.
c
contribution.
unif.
x
is known in closed form (25), whereas the
correlation contribution
unif.
c
is not. The versions of the LDAin common use (2628)
are t to numerical values of
unif.
c
derived fromQMCcalculations for the unpolarized
( = 0) and fully polarized ( = 1) uniformelectron gas (29). They differ in the details
of the functional forms employed for the ts and in the modeling of the -dependence
in
unif.
c
. Despite its apparent simplicity, the success of the LDA can be attributed to
the fact that it satises sum rules for the exact exchange and correlation holes (3032)
and provides a good description of the on-top hole density (33).
The GGA considers slowly varying deviations from the uniform electron gas and
is a functional of both the density and its gradient,
E
GGA
xc
[
, |
|, |
|] =
_
dr(r)
GGA
xc
[
(r),
(r), |
(r)|, |
(r)|]. (7)
264 Huang
Carter
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
Although sometimes referred to as nonlocal, the GGA is semilocal in the sense that it
depends on both the value of the density and its gradient at r. Note that the GGAdoes
not represent a systematic improvement over the LDA, and in some cases it can give
worse agreement with experiment. Unfortunately, there is no unique way to specify
GGA
xc
, so a greater variety of GGA functionals is in common usage compared to the
LDA. The Perdew and Wang (PW91) functional (3436) is constructed to satisfy a
number of known limits and bounds and is nonempirical. The related Perdew, Burke,
and Ernzerhof (PBE) functional (37, 38) represents a simplication of PW91. Zhang
&Yang proposed a revised version of PBE(revPBE) (39, 40) by replacing the value of
one parameter in PBE with a different value obtained by tting to a set of theoretical
atomic energies, improving its performance for atoms and molecules. An alternative
revision of PBE by Hammer, Hansen, and Nrskov (RPBE) (41) yields results com-
parable to revPBE but does not involve t parameters. Benchmark studies indicate
that both revPBE and RPBE can provide an improved description of chemisorption
compared to PW91 and PBE (41).
There is a great deal of accumulated experience with the LDA and the GGA, and
from this a few well-known trends are apparent (2, 35, 36, 4244). Both the LDA and
the GGA give good ground-state structural properties (e.g., lattice parameters) to
within a few percent of experiment. The LDA tends to overbind: Lattice parameters
are underestimated and bulk moduli are too high. The GGA usually (but not always)
improves on the LDA structural properties, with a slight tendency to underbind, al-
though the magnitude of the GGAerror is typically not as large as the LDA. Cohesive
energies are overestimated in the LDA, and this error can be signicant. The GGA
usually improves on this. Errors in the cohesive energies can often be traced back to a
poor description of the isolated reference atom. Band gaps fromthe LDA are way too
small; for example, the LDA gap for Si is approximately 50%of experiment (45). The
GGA does not help. Although association of the DFT band gap with the experimen-
tal gap is not justied, underestimation of the gap in semiconductors has unfortunate
consequences even for the ground state: The conduction band can overlap with the
valence band, incorrectly predicting a metallic ground state (e.g., for Ge) (46).
In the case of only one electron, the KS total energy in Equation 1 should simply
reduce to a sum of the noninteracting kinetic energy T
s
and the electron-ion energy.
That is, the exact E
xc
should completely cancel out J when N = 1. This is true in HF
theory, in which the nonlocal Fock exchange,
E
HF
x
=
1
2
occ.
i, j
_
dr dr
i
(r)
j
(r)
j
(r
)
i
(r
)
|r r
|
, (8)
takes the place of E
xc
. But the local and semilocal functionals in standard usage do not
exhibit this exact cancellation; the spurious contribution that remains is referred to
as the self-interaction error. This error tends to be more severe for strongly localized
densities, e.g., the d- or f-electrons in late transition metal oxides, in which the self-
interaction error tends to favor electron delocalization, leading to a metallic ground
state when the correct ground state should be insulating.
Although not in common use for solids, we also mention the meta-GGAs (42, 47).
In addition to the usual GGA inputs, these also require the Laplacian of the density,
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
,
2
, and the local kinetic energy density,

(r) =
1
2
occ.
j
(r)
2
. (9)
2.2. Hybrid Functionals
One way to minimize the self-interaction error is to simply add back some amount
of Fock exchange. The functionals that do this are known as hybrids, and as with the
GGAs, there is no unique way to construct a hybrid functional out of E
HF
x
. Instead,
the hybrid functionals popular in molecular applications (e.g., B3LYP) take a more
pragmatic approach, by introducing parameters (sometimes many) that typically are
t to thermochemical data to achieve chemical accuracy. The simplest of the hybrid
functionals in use for solids is the PBE0 functional (48),
E
PBE0
xc
= E
PBE
xc
+a
_
E
HF
x
E
PBE
x
_
= a E
HF
x
+(1 a)E
PBE
x
+ E
PBE
c
, (10)
where E
PBE
x
and E
PBE
c
are the exchange and correlation components of the semilocal
PBE functional, respectively. A mixing factor of a = 1/4 is argued from theoretical
grounds and not derived from ts, although other empirical choices for a have been
explored. Although hybrid functionals can deliver better agreement with experiment
compared to the LDA/GGA, evaluation of the nonlocal exchange in periodic solids
is more computationally expensive compared to the LDA/GGA. Nevertheless, the
use of nonlocal Fock exchange with periodic boundary conditions is now available in
a number of electronic structure packages (e.g., GAUSSIAN03, CRYSTAL).
For zero-band-gap systems (metals), E
HF
x
is long-ranged, which presents addi-
tional difculties in its evaluation. This is also the reason why HF theory for metals
yields (k)/k that diverge logarithmically for states k near the Fermi level. It is
known that the way to eliminate this divergence is to screen the Coulomb poten-
tial 1/|r r
| appearing in the exchange operator. In the Thomas-Fermi theory of

screening, this is done by replacing 1/|rr
| in E
HF
x
with e
k
TF
|rr
|
/|rr
|, where k
TF
is the Thomas-Fermi screening length. In the many-body Greens function theory,
the Coulomb potential appearing in the self-energy operator is screened by the
inverse dielectric function
1
(see Section 3). Bylander & Kleinman (49) proposed
a hybrid functional that incorporates Thomas-Fermi screening and found band gaps
for Si much improved over the LDA. Seidl et al. (50) later argued for the validity
of this procedure using a constrained-search formulation of DFT and showed that,
in principle, the DFT band gaps derived from such a screened exchange operator
can include some amount of the derivative discontinuity . More recently, Heyd and
coworkers (51, 52) proposed a screened hybrid functional based on PBE0, derived by
partitioning the Coulomb operator into short- and long-range contributions, whose
relative extents are controlled by an adjustable parameter .
The performance of hybrid functionals in solids is not as well-established as the
LDA/GGA. Probably the most signicant improvement provided by the hybrid func-
tionals is in the DFT band gaps. In particular, the narrow-gap semiconductors that
are metals in the LDA/GGA are correctly predicted to be semiconductors (50, 53).
266 Huang
Carter
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
Band gaps of large-gap insulators are still underestimated. The performance for met-
als is not as satisfactory, as cohesive energies are worse than those derived from PBE,
and valence bandwidths are too large (54, 55).
2.3. The DFT + U Method
Another strategy to deal with the self-interaction error is the DFT + U method,
commonly used to model strongly correlated materials (56). Here an additional (U
J ) parameter is added to the total DFT energy functional to mimic an effective on-
site Coulomb (U) and exchange (J ) between electrons with the same orbital angular
momentum. Such an HF-like correction is motivated by the fact that HFtheory is free
of self-interaction errors. One advantage of DFT + U over the hybrid functionals
described above is that the additional computational expense is minimal, whereas the
evaluation of nonlocal Fock exchange in a periodic solid is much more demanding.
Although one can choose the (U J ) parameter empirically, a more rigorous
approach is to determine (U J ) from rst principles. One way to do this is via
constrained DFT (57), in which the electron occupation on a particular site is held
xed. One can then extract effective U and J values from the change in total energy
as a function of the number of xed electrons. Drawbacks are that (U J ) is derived
froman unphysically constrained situation, and the underlying DFTcalculations still
suffer from approximate XC problems.
More recently, Mosey & Carter (58) proposed an alternative approach for esti-
mating (U J ). They modeled chromia (Cr
2
O
3
) as clusters embedded in a classical
electrostatic environment (see Section 6.1) and explicitly derived effective U and J
parameters from unrestricted HF theory, which is strictly self-interaction-free. The
resulting parameters were somewhat higher than that obtained from previous con-
strained DFT estimates, but their subsequent use in a DFT + U calculation gave
results consistent with experiments.
An extension of the DFT+U method is dynamical mean-eld theory (11), which
goes beyond the on-site HF description of the correlated electrons. As a result, situ-
ations such as metal-insulator transitions can be treated consistently within the same
theory. However, dynamical mean-eld theory still assumes a Hubbard-Anderson
form for the correlation, which ultimately could constrain the resulting physics.
3. QUASIPARTICLE EXCITATIONS
In a photoemission experiment, a photon impinges on a solid and ejects an electron.
Inverse photoemission is essentially the time reversal of photoemission: An electron
is absorbed by a solid, leading to the emission of a photon. The spectra obtained
from such experiments correspond to excitations between N- and (N 1)-electron
states, also known as quasiparticle excitations. We discuss in this section a route
to quasiparticle excitations, which is grounded in the theory of many-body Greens
functions. The approach is based on the GW approximation, originally formulated
by Hedin (59) in 1965 for the uniform electron gas and later applied to realistic bulk
semiconductors in the mid-1980s (24, 60). Today, calculations involving GW method
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
are considered routine and can yield accuracy on the order of 10
1
eV of experiment
for the band structure of bulk semiconductors (46).
3.1. Basic Theory
In the HF theory for noninteracting electrons, the spectral representation of the one-
electron Greens function G
0
has poles that occur at the HF eigenvalues
j
, which,
according to Koopmans theorem, correspond to one-electron removal/addition en-
ergies. Thus, G
0
provides a natural way to describe photoemission and inverse pho-
toemission processes. However, the noninteracting Greens function G
0
does not
account for orbital relaxation effects; i.e., the electrons of the bulk sample should
respond to screen the additional hole or particle added to the system. A proper
description of screening needs to take into account many-body effects. For an inter-
acting many-electron system, the one-particle Greens function G can be related to
the noninteracting Greens function G
0
via the Dyson equation,
G() = G
0
() + G
0
()()G(), (11)
where is referred to as the self-energy, which encapsulates the complicated effects
of XC.
In the presence of interactions, the spectrum associated with G is no longer guar-
anteed to be the spectrum of one-electron removal/addition energies associated with
G
0
. Nevertheless, if the spectrumcan be regarded as a superposition of sharply peaked
functions of the energy, it is meaningful to speak of quasiparticle excitations, which
are specied by
[

T +
ion
(r) +
J
(r)]
QP
j
(r) +
_
dr
(r, r
;
QP
j
)
QP
j
(r
) =
QP
j

QP
j
(r). (12)
Here,
QP
j
and
QP
j
are quasiparticle wave functions and energies, respectively. Note
the correspondence with HF theory. The HF exchange potential is now replaced
by , which is nonlocal, non-Hermitian, and energy dependent. Consequently, the
quasiparticle wave functions are nonorthogonal, and the energies are complex, with
lifetimes determined by the imaginary components. In analogy with HF theory, the
quasiparticle wave functions can be identied as effective single-particle (hole or
electron) wave functions, whereas the quasiparticle energies correspond to energies
associated with the removal or addition of an electron.
The basic task is to obtain an approximation for . A practical strategy was pro-
vided by Hedin (59), who developed a perturbative expansion for in terms of a
dynamically screened interaction W(r, r
; ). Physically, W represents the potential

at r due to a test charge at r
, taking into account the polarization of the surround-

ing electrons. For a material with a large polarization, this quantity should be much
weaker than the bare Coulomb potential 1/|rr
|, and thus we expect an expansion in

W to behave better than an expansion in 1/|rr
|. The leading term in the expansion

of in powers of W is
(r, r
; ) =
i
2
_
d
G(r, r
; +
)W(r, r
). (13)
268 Huang
Carter
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
The GW approximation refers to the use of Equation 13 above for and the neglect
of higher-order terms in the expansion (vertex corrections). Here, W is the bare
Coulomb potential screened by the dielectric function ,
W(r, r
; ) =
_
dr
1
(r, r
; )
1
|r
|
, (14)
where is given by
(r, r
; ) = (r r
)
_
dr
P(r
, r
; )
1
|r r
|
, (15)
and P is the polarization function. Within the random phase approximation, which
describes the mean-eld response to the total potential, this is
P(r, r
; ) =
i j
( f
i
f
j
)
i
(r)
j
(r)
j
(r
)
i
(r
)
(
j

i
) +i
, (16)
where
i
and
i
are mean-eld wave functions and energies, respectively, with occu-
pations f
i
.
3.2. Standard Implementation
Many technical issues arise when applying the GWmethod to realistic materials. Most
GW band-structure calculations begin with KS-DFT within the LDA for XC. This
yields a set of KS wave functions
nk
and energies
nk
, where n is a band index, and
k lies in the rst Brillouin zone. The quasiparticle equation (Equation 12) becomes
_

T +
ion
(r) +
J
(r) +
LDA
xc
(r)
_
QP
nk
(r)
+
_
dr
_
r, r
;
QP
nk
_
QP
nk
(r
) =
QP
nk

QP
nk
(r),
(17)
where
LDA
xc
is the LDA XC potential, which is local and independent of the energy,
and
_
r, r
;
QP
nk
_
=
_
r, r
;
QP
nk
_
LDA
xc
(r). (18)
The reason for starting with the LDA is that if
LDA
xc
is a good approximation for the
true XC potential, then the difference should be small, and the LDA wave func-
tions
nk
are close to the quasiparticle wave functions
QP
nk
. Initial GW applications
to simple, bulk sp-bonded semiconductors found that the overlap
nk
|
QP
nk
is better
than 99% (61). Note that more recent work involving surfaces (62) and nite-sized
systems (6365) does nd appreciable differences between the KS and quasiparticle
wave functions.
An approximation for the self-energy (Equation 13) is constructed starting with
the noninteracting Greens function G
0
and the dynamically screened interaction W
(Equations 1416), both evaluated using the KS wave functions
nk
and energies
nk
.
This is referred to as = G
0
W
0
. These quantities are typically evaluated in reciprocal
space, although real-space implementations are possible. The most straightforward
way to accomplish the frequency-domain convolution for is via a numerical inte-
gration over the frequency (24, 66), which requires a time-consuming construction
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
and inversion of (k, ) at each value of in the quadrature. Alternatively, model
dielectric functions have been proposed (61, 67, 68) that provide a functional form
for the frequency dependence of
1
(k, ) so that the frequency integration can be
done analytically. They can work reasonably well for sp-bonded materials but are less
successful for materials with localized d- and f-electrons. More recent work evaluates
W and without resorting to model dielectric functions, either by working in the
space-time domain (69) or in the wave vectorfrequency domain (7072).
The Dyson equation (Equation 11) relates G
0
to the interacting Greens function
G via . In principle, starting with an initial = G
0
W
0
, one should self-consistently
solve the Hedins equations (59) for the operator quantities G, , P, and W. In
practice, however, this is not done; most studies stop at = G
0
W
0
and do not
explicitly solve Equation 17. Instead, the quasiparticle energies
QP
nk
are obtained
by perturbatively correcting the KS energies
nk
to rst order in , in the basis
of the KS wave functions
nk
. Whereas early reports on simple, bulk sp-bonded
semiconductors indicate that second-order perturbation corrections are small (24),
later work showed that the quality of the quasiparticle energies is tied to the quality
of the input KS quantities (73, 74).
Although the non-self-consistent G
0
W
0
approximation has the obvious advan-
tage of requiring much less numerical effort than a fully self-consistent solution, a
non-self-consistent G leads to a formal violation of some important sum rules (e.g.,
particle conservation). The role of self-consistency in the GW method and how it
should be imposed are ongoing topics of discussion (7582). A series of studies (83
86) involving the homogeneous electron gas with non-self-consistent G
0
W
0
, partially
self-consistent GW
0
, and fully self-consistent GW found that the description of the
quasiparticle spectrum deteriorates with increasing levels of self-consistency. Yet the
fully self-consistent GW yields total energies for the homogeneous electron gas that
agree well with QMC results. For the purpose of band-structure calculations, it ap-
pears that the non-self-consistent G
0
W
0
approximation is preferred.
A second question concerns the validity of the pseudopotential approximation.
Most GW work to date relies on the use of pseudopotentials to represent core elec-
trons, which neglects the core-valence contribution to the self-energy . More re-
cently, GWhas been implemented within the various all-electron schemes: linearized
augmented plane wave (LAPW) (66), full-potential linearized augmented plane wave
(FLAPW) (78, 87), full-potential linear mufn-tin orbital (LMTO) (74, 88), and pro-
jector augmented wave (PAW) (71, 72). Although a number of technical differences
exist between the different all-electron methodologies, they do appear to consistently
underestimate band gaps relative to pseudopotential-based GW calculations. While
it has been argued that the errors due to the lack of self-consistency cancel the errors
due to the neglect of vertex corrections (79), the body of all-electron work suggests
that the signicance of pseudopotential errors has not been fully appreciated.
4. NEUTRAL EXCITATIONS
Excitations observed in optical spectroscopy and EELS (electron energy loss spec-
troscopy) do not change particle number and are fundamentally distinct from the
270 Huang
Carter
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
BSE: Bethe-Salpeter
equation
quasiparticle excitations discussed in Section 3. These single neutral excitations
can be regarded as a two-(quasi)particle excitation involving a (quasi)electron and
a (quasi)hole, which are allowed to interact. This information is contained in the
poles of the two-particle Greens function (6), and the equation whose solution cor-
responds to these poles is known as the Bethe-Salpeter equation (BSE). The BSE is
closely related to the linear-response formulation of TDDFT (7, 8991, 92, 9395),
which is in widespread use for the optical excitations of molecules and nite clusters.
A comparison of the BSEand TDDFThas been thoroughly discussed by Onida et al.
(6).
The excitonic effects due to the interaction of a particle and hole can be described
as a mixing of particle-hole pair states
ph
c
ph
| ph, where | ph is an antisymmetrized
product containing a particle in state p and a hole in state h. The particle-hole ampli-
tudes c
ph
and excitation energies E are obtained from the BSE, which one can write
as an eigenvalue equation (96, 97)
__
QP
p

QP
h
_
E
_
c
ph
+
,h
K
ph, p
h
c
p
h
= 0, (19)
where
QP
p
,
QP
h
are quasiparticle energies, and the kernel
K
ph, p
h
= K
Coul.
ph, p
h
+ K
dir.
ph, p
h
. (20)
The rst term above is the Coulomb (also known as exchange) kernel involving the
bare Coulomb interaction,
K
Coul.
ph, p
h
= 2
_
dr dr
p
(r)
h
(r)
1
|r r
p
(r
h
(r
), (21)
and the second termis known as the direct kernel and involves a Coulomb interaction
screened by the inverse dielectric function
1
(evaluated at = 0),
K
dir.
ph, p
h
=
_
dr dr
p
(r)
h
(r
)
1
(r, r
)
1
|r r
p
(r)
h
(r
). (22)
In practice, the procedure involves (a) a KS-DFT calculation to obtain a set of KS
orbitals
j
and eigenvalues
j
; (b) a GW calculation (Section 3) for the quasiparticle
energies
QP
j
, using the KS quantities as input; and (c) the solution of Equations 1922
using
QP
j
from GW and
j
from KS-DFT.
The linear-response formulation of TDDFT leads to a set of equations for the
excitation energies similar to Equations 1922 above, except the direct kernel K
dir.
ph, p
is replaced by the TDDFT XC kernel

K
xc
ph, p
h
=
_
dr dr
p
(r)
h
(r
)

2
E
xc
(r)(r
p
(r)
h
(r
). (23)
Due to the local nature of most approximations for E
xc
, the two-electron integral
over r and r
above reduces to a one-electron integral over r. The widespread use

of TDDFT for molecular and cluster problems largely results from the fact that
it requires relatively little effort compared with other excited-state wave function
based methods and it is available in many quantum chemistry packages. However,
the application of TDDFT to periodic systems is problematic (6, 91, 93), providing
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
little improvement over KS energies, and the reasons for this can be traced back to
the locality of E
xc
. One line of work to overcome these issues with solids is to go to a
time-dependent current density functional theory using the functional of Vignale &
Kohn (98), in which the current density replaces the density as the central quantity
of interest. However, the general applicability of the Vignale-Kohn functional is still
under debate (99).
5. QUANTUM MONTE CARLO
The QMC method (810, 100) is fundamentally different from the other quantum
many-body techniques discussed in this review. Here one computes the expectation
value of an observable

O by stochastically sampling froma distribution function given
by the full, 3N-dimensional wave function . Unlike other wave functionbased
methods, is not restricted to the usual Slater determinantal form, and thus one can
incorporate more efcient ways of describing electron correlation. QMC is a general
means of treating electron correlation in periodic crystals, beyond that provided by
DFT with current XC functionals, and in some cases, specic excited states can also
be accessed. An attractive feature of QMC is that it scales as O(N
3
), similar to KS-
DFT (although the prefactor is considerably larger), and is easily parallelized. O(N)
implementations have also been reported (101, 102). Other methods that include
comparable levels of electron correlation effects or treat excited states (e.g., correlated
quantum chemistry techniques) typically exhibit far worse scaling and can be difcult
to apply to periodic systems. Most successful applications of QMC to the solid state
have been for covalently bonded semiconductors and insulators, for example, C and
Si nanoclusters and their bulk counterparts (103108).
Two types of QMC methodologies are in common use for solids. The rst is
variational Monte Carlo (VMC), which refers to the Monte Carlo optimization of a
many-body trial function
T
and the evaluation of expectation values of an observable
O with respect to
T
. Here, the choice of
T
is not restricted to forms that are easily
integrated, e.g., Slater determinantal wave functions. The only requirement is that
we have to be able to evaluate
T
and the action of

Hon
T
. QMCtrial functions are
usually of the Slater-Jastrow form, in which electron correlation effects are included
by multiplying a Slater determinantal function by a factor e
J
, where J is a Jastrow
correlation function. J consists of a sum of one- and two-electron terms,
J(r
1
, . . . , r
N
) =
N
i =1
t
1
(r
i
) +
N
i <j
t
2
(r
i
, r
j
). (24)
Details of Slater-Jastrow functions are discussed further in Reference 9. The Slater
determinantal part of
T
is taken from a standard HF or KS-DFT calculation, and
the VMC optimization process then varies the parameters in
T
to minimize some
estimator; this can be the energy, the variance of the energy, or some combination of
both.
The main utility of VMC is to obtain an optimized
T
for use in a subsequent
xed-node diffusion Monte Carlo (DMC) calculation. Fixed-node DMC provides
272 Huang
Carter
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
expectation values of the form

O =

T
|

O |
0
T
|
0
, (25)
where
0
is the wave functionthat minimizes the total energy, subject tothe constraint
imposed by the nodes of
T
. If the nodes of
T
correspond to the nodes of the exact
ground state, then
0
will be the exact ground state, regardless of the details of
T
away from the nodes. In this case,

O is exact if

O is the Hamiltonian

H or if
O commutes with

H. The many-electron wave function has nodes due to fermion
antisymmetry, and in general these nodes are complicated and not well understood.
Most DMC applications employ a
T
with an approximate nodal structure, and
the use of this is known as the xed-node approximation. In practice, Slater-Jastrow
trial functions provide a reasonable description of the nodal structure, and xed-
node DMC can recover 95% of the correlation energy (109). The availability of
a high-quality
T
is critical to the success of DMC in at least two ways: (a) The
nodal structure of
T
determines the accuracy of the xed-node approximation, and
(b)
T
serves as a guiding function for importance sampling, and thus it controls the
efciency of the algorithm.
A number of implementation issues arise in the application of QMC to solids.
First, near the atomic core regions, the wave function is rapidly varying, and thus
it is difcult to sample the wave function efciently and accurately in these regions.
Although this can be somewhat ameliorated with the use of pseudopotentials (110),
transition metals pose a problem, as the strongly localized d- and f-electrons have to
be treated as valence. To date, only a few QMC studies of solids containing transition
metals have been reported, e.g., transition metal oxides (111, 112). Nonlocal pseu-
dopotentials in DMC need to be localized in an approximate manner; this introduces
additional complications, which can be minimized by using an accurate trial func-
tion (113). Second, QMC with periodic boundary conditions suffers from nite-size
effects not seen in mean-eld methods such as KS-DFT. Unlike mean-eld meth-
ods, electrons in QMC are explicitly correlated in real space, and thus the problem
cannot be reduced to a single real-space primitive cell via Blochs theorem. Periodic
QMC calculations are done in a large supercell, and steps need to be taken to correct
for nite-size errors. Finally, it is difcult to obtain accurate forces, as the QMC
Hellmann-Feynman estimator gives an innite variance. Other attempts at forces in-
clude nite differencing, using correlated sampling techniques to reduce the variance
(114), and adding Pulay corrections to the Hellmann-Feynman forces (115), to name
a few.
6. EMBEDDED CLUSTER METHODS
Many solid-state phenomena of interest involve a single, localized inhomogeneity in
an extended background (e.g., vacancies, impurities, adsorbates). A natural strategy
in this case is to focus the modeling effort in the neighborhood around the inhomo-
geneity. Typically, this is a small, embedded cluster consisting of the inhomogeneity
and some of the nearby atoms, while the effects of the background are incorporated
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
in a more approximate manner. The difculty with embedding methods is that they
tend to be highly system specic, as there is no unique way to dene the cluster-
background partitioning and their coupling interaction. The possibility of artifacts
due to nite-size effects is always a concern, and computed properties often do not
converge smoothly as a function of cluster size and morphology.
Although periodic supercell models are easy to dene and converge for supercell
sizes up to 10
2
atoms, one can envision cases in which supercell models become
impractical and embedded cluster models provide a useful alternative. Moreover, local
renement of the physical model becomes possible in such schemes, e.g., improved
descriptions of quantum mechanical effects in a local region.
Bare and embedded cluster models for the solid state have a long history, particu-
larly in the study of adsorbates on surfaces, and many different embedding strategies
have been proposed (116119). Below we review four classes of embedded cluster
schemes: purely electrostatic, wave function partitioning, density based, and hybrid
quantum/classical force eld approaches.
6.1. Electrostatic Embedding
Ionic crystals possess relatively localized electron densities, and thus small clusters
can provide a reasonable description of local phenomena, provided that the surround-
ing long-ranged electrostatics are accounted for properly. The simplest embedding
models involve a small cluster of ions embedded in a set of point charges (typically
several thousand) representing the background. The electronic structure of the em-
bedded cluster is solved using standard, real-space quantum chemistry methods, and
thus the point-charge lattice is necessarily nite. Unfortunately, the long-range na-
ture of the Coulomb potential means that the Madelung potential within the cluster
converges slowly in real space with respect to the number of point charges. If this is
not specied carefully, one can get a Madelung potential that disagrees with standard
Ewald summation techniques. The easiest way to dene the charges is Evjens (120)
method. For a perfect cubic lattice, formal ionic charges are placed at lattice positions
within a large box around the cluster, and fractional charges are assigned to sites at
vertices (1/8), edges (1/4), and faces (1/2). More sophisticated procedures employ
the Ewald summation to compute the exact Madelung potential at various sampling
points throughout the solid and then t the background charges to these values (121,
122).
The presence of a bare Coulomb potential at sites that are nearest neighbors of the
embedded cluster can articially distort the wave function at the cluster boundary.
One can minimize this edge effect by putting an effective core potential at these
positions (123) to mimic the Pauli repulsion due to electrons on the nearest-neighbor
ions. An extension of this is to embed the cluster in a set of ab initio model potentials
(AIMPs) (124), in addition to the usual point-charge array representing the Madelung
potential. Typically several hundred AIMP sites are included. In short, the AIMP can
be viewed as a type of HF effective core potential containing (a) long-range and (b)
short-range Coulombic potentials, (c) an exchange potential, and (d ) a projection
operator to enforce core-valence orthogonality.
274 Huang
Carter
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
CI: conguration
interaction
6.2. Orbital Space Partitioning
Orbital space partitioning begins with a mean-eld HF or KS-DFT description for
the total system and then subsequently partitions the space of one-electron orbitals.
The orbitals from an HF or KS-DFT calculation are typically delocalized over the
entire system. However, because a single Slater determinant wave function is invari-
ant under a unitary transformation of the orbital basis, the key step is to nd an
appropriate transformation that localizes a subspace of orbitals on the predened
embedded region, and a complementary subspace on the background. Many ways
have been proposed to do this (e.g., 125).
AGreens functionformulationof the embedding problemcanbe tracedas far back
as the work of Pisani and coworkers (126130) in the 1970s. These methods begin
by rst nding a mean-eld solution for the total, defect-free system (e.g., the clean
substrate, in the case of chemisorption). These orbitals are subsequently transformed
into a set localized in the embedded region and a set localized in the background. The
Greens function for the total system with the defect is solved in terms of the Greens
function of the background, whose matrix elements are assumed to be those from the
defect-free system in the basis of localized background orbitals. A related approach
for surface problems is the surface Greens function methods (131, 132), which yield
a proper description of surface states, unlike models involving periodic slabs of nite
thickness. The surface Greens function approaches consider a few surface layers in
the presence of a DFTeffective potential derived fromthe bulk solid, whose role is to
force the surface wave functions to match the electronic structure of the bulk region.
These Greens functionbased approaches all involve a mean-eld description of
the embedded region and do not easily allowfor the explicit introduction of correlated
wave functions. To this end, Whitten and coworkers (133) developed an embedded
conguration interaction (CI) method. In short, the procedure involves (a) nding
an HF solution for a large cluster surrounding the defect of interest, (b) localizing
orbitals in or near a smaller, embedded region within the large cluster, (c) constructing
effective Coulomb and exchange operators out of these localized orbitals, (d ) aug-
menting the basis set within the embedded region, and (e) performing a CI calculation
using orbitals localized in the embedded region. Thus, the presence of electrons out-
side the embedded region is accounted for approximately; however, the background
environment is taken to be a larger cluster, which may not reect accurately a true
bulk environment.
6.3. Density-Based Embedding
Orbital-based embedding schemes that rely on localizing orbitals to span a nite,
embedded subregion are difcult to apply to metals because the conduction elec-
trons in a metal are delocalized by nature (134). Alternatively, one can formulate
the embedding in terms of the density. The essential idea behind this class of em-
bedding schemes is to begin with a DFT-based description and to model the un-
known subsystem-environment interaction with a known, but approximate, density
functional.
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
The development of density-based embedding can be traced as far back as Cortona
(135), who sought a reduced-scaling DFT method for the cohesive energies of bulk
solids. This work, and much of the subsequent work based on it (136150), can be
understood as a partitioning of the total system density
tot
into an embedded region
of interest
I
, and a background
II
, such that
tot
=
I
+
II
. A KS solution for the
embedded region (I) is given by
_
1
2
2
+
s
[
I
](r) +
emb
[
tot
,
I
](r)
_
I, j
(r) =
j
I, j
(r), (26)
where
s
is the usual KS potential for region I, and
emb
is an effective, one-electron
potential representing the embedded region/background coupling and has the form
emb
[
tot
,
I
](r) =
Ts
[
tot
,
I
](r) +
J
[
tot
,
I
](r) +
xc
[
tot
,
I
](r) +
tot
ion
(r)
I
ion
(r).
(27)
Here,
Ts
,
J
, and
xc
are the kinetic, Hartree, and XC contributions, respectively:
Ts
[
tot
,
I
](r) =
T
s
[
tot
]
tot
(r)

T
s
[
I
]
I
(r)
, (28)
J
[
tot
,
I
](r) =
J[
tot
]
tot
(r)

J[
I
]
I
(r)
, (29)
xc
[
tot
,
I
](r) =
E
xc
[
tot
]
tot
(r)

E
xc
[
I
]
I
(r)
. (30)
Unlike region I, the kinetic energy contribution to the embedded region/background
coupling is modeled with an approximate T
s
, which is an explicit functional of the den-
sity. Cortonas original work employed the Thomas-Fermi model (1) for T
s
, although
other approximations can be used as well. Thus, the determination of the density of
the total system is recast into a problem involving a smaller region of interest in an
embedding potential,
emb
. The embedding potential is an explicit functional of the
total density
tot
and the subsystem density
I
, and is solved for self-consistently.
Wesolowski & Warshel (136) applied a similar density-based embedding to the
study of single-molecule solvation. Although the subject of liquid-state simulations
is beyond the scope of this review, their methodology is general and relevant to our
discussion of the solid state. The total system density
tot
=
I
+
II
is partitioned
into the electron density of the embedded solute,
I
, and the electron density of the
surrounding solvent,
II
. The electron density of each solvent molecule is determined
individually in advance and frozen at this initial value, and
II
is constructed as the sum
over all solvent molecules. Cortonas approach is followed in an analogous manner, to
arrive at a set of KS equations whose structure is identical to Equations 2630, except
now
I, j
are the orbitals of the solute molecule, and the subscripts I and tot refer to the
solute molecule and the combined solute/solvent system, respectively. More recently,
this approach has been extended to include a TDDFT treatment of the embedded
solute molecule to evaluate solvent-induced spectral shifts (137). Clearly, the use of
a model T
s
is a signicant approximation in the theory, and various approximations
to T
s
in the context of atomic and molecular systems have been explored (138).
276 Huang
Carter
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
All the density-based embedding applications described above represent a DFT-
in-DFT embedding, in the sense that they begin with a DFT estimate for the back-
ground region
II
and solve for the embedded DFT density
I
in the presence of
II
. Carter and coworkers (141149) developed a related density-based embedding
method that includes electron correlation within the embedded region beyond the
DFT level to address problems ranging from chemisorption to strongly correlated
Kondo physics. The derivation of the working set of embedding equations differs
somewhat from Cortonas approach, although the structure of the resulting embed-
ding potential is the same. Cortonas derivation involves the functional minimization
of the total DFT energy with respect to the set of one-electron orbitals
I, j
span-
ning the embedded region (I). To move beyond a one-electron picture, Carter and
coworkers began by writing the total energy functional as
E
tot
[
tot
] = E
I
[
I
] + E
II
[
II
] + E
int
. (31)
E
int
is a generally unknown cluster-background interaction and is modeled as
E
int
E
DFT
tot
[
tot
] E
DFT
I
[
I
] E
DFT
II
[
II
], (32)
where E
DFT
i
(i = tot, I, II) are known, within the context of approximate density
functionals,
E
DFT
i
[
i
] = T
s
[
i
] + E
xc
[
i
] +J[
i
] +
_
dr
i
ion
(r)
i
(r). (33)
We emphasize that no approximation is introduced for E
I
and E
II
; only the cluster-
background coupling E
int
is approximated with a model DFT functional. Instead of
searching for a mean-eldsolutionthat minimizes a DFTtotal energy, one considers a
more general variationwithrespect tothe embeddedregiondensity
I
. If the inuence
of the inhomogeneity is sufciently contained within the embedded region I, and the
background region II is large, it is reasonable to assume that E
II
[
II
]/
I
= 0. This
leads to the Euler equation
E
tot
[
tot
]
I
(r)
=
E
I
[
I
]
I
(r)
+
emb
[
I
,
tot
](r) = , (34)
where is a Lagrange multiplier associated with the constraint
_
dr
I
(r) = N
I
, and
emb
is an effective embedding potential,
emb
[
I
,
tot
](r) =
E
DFT
tot
[
tot
]
tot
(r)

E
DFT
I
[
I
]
I
(r)
. (35)
Substituting Equation 33 into E
tot
and E
I
above leads to an embedding potential
emb
identical to that given by Equations 2730.
The density
I
, which satises Equation 34, equivalently can be found by solving
the Schr odinger equation for the embedded cluster,
_

H
I
+
N
j =1
emb
[
I
,
tot
] (r
j
)
_
I
= E
I
, (36)
where

H
I
is a many-body Hamiltonian involving the ions and electrons of the em-
bedded cluster region, and
I
is a many-body wave function from which
I
can be
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
derived. At this point, the stage is set for the application of well-established ab initio
methods (e.g., CI) to systematically include electron correlation within the embedded
cluster region.
Of the quantities
tot
,
I
, and
II
, only two are independent, so in principle one
should minimize the energy with respect to the two chosen independent variables,
subject tothe constraint that
tot
=
I
+
II
. Sofar this has not beenpossible inpractice,
and a number of approximate schemes have been tried. Early efforts (141145) chose
to begin with a
tot
obtained from a periodic KS-DFT calculation. The intention
was to then input this
tot
into the embedding potential
emb
and solve Equation 36
self-consistently for
I
. However, numerical difculties arose in the self-consistent
update of
emb
, so the kinetic energy contributions T
s
[
I
]/
I
and T
s
[
tot
]/
tot
were
held xed at some initial estimates (frombare cluster and periodic DFTdensities) and
not solved for self-consistently. Later work (146, 147) adopted the strategy taken in
the DFT-in-DFT approaches described above, by explicitly constructing an estimate
for the background density
II
and then holding this quantity xed. Equation 36 is
then solved self-consistently for
I
, with
tot
evaluated as a sum of the embedded
cluster density
I
and the initial estimate for the background density
II
. This led to
a much more numerically stable procedure, allowing all terms in
emb
to be solved
self-consistently. Lahav & Kl uner (150) also recently reported an implementation of
a similar self-consistent embedding scheme.
6.4. Solid-State QM/MM
QM/MMrefers to a class of hybrid methods in which an embedded region of interest
is treated quantum mechanically (QM), while the background region is described
using classical force eld models [molecular mechanics (MM)] (151). This is a popular
strategy in biomolecular simulations, and an increasing effort has been made over the
past decade to apply these ideas to the solid state. The motivation is to allow for the
embedded structure to relax while accounting for the mechanical and polarization
response of the background, and models that attempt to do this can become quite
complex. There are two general issues with which all QM/MM approaches need to
cope: (a) howto dene the QM/MMboundary and (b) howto deal with the mismatch
of forces across the QMand MMregions. The nonuniqueness of the solution to these
two problems has spawned a myriad of QM/MM-type methods.
In an ionic or molecular solid, there is a natural separation between ions or
molecules, and interatomic interactions are largely dominated by classical electrostat-
ics. Thus, the specication of the QM/MM boundary is relatively straightforward.
An important consideration is a proper treatment of long-range polarization effects,
which is typically done for ionic materials via shell model potentials for the back-
ground ions (152). In the shell model (153), each ion is assigned an effective positive
core and a negative shell that are connected via a harmonic potential, and polar-
ization manifests as a redistribution of the effective charges. The AIMP embedding
(Section 6.1) has also been combined with shell models in a self-consistent manner
(154). More recently, R osch and coworkers (155) developed an elastic polarizable
environment model to study the adsorption of a Pd atom at oxygen vacancies of
278 Huang
Carter
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
MgO(001), in which the background is described with effective core potentials, point
charges, shell models, a dielectric continuum, and special short-range potentials de-
signed to handle the embedded cluster/background boundary region.
Covalently bonded materials present additional problems because it is not so obvi-
ous howto determine the QM/MMpartitioning. One common approach is to choose
the QM/MMboundary to cut across covalent bonds, which leaves dangling bonds as-
sociated with undercoordinated boundary atoms. These dangling bonds are typically
saturated with a ctitious pseudoatom (116), a strategy that follows from early efforts
by Goddard and coworkers (156) to study the reactivity of Si surfaces with nite Si
cluster models. Examples of this include the studies of Carmer et al. of the reaction
of acetylene on the Si(111) surface (157); the QM-Pot method of Sauer & Sierka
(158); and the SIMOMM (surface-integrated molecular orbital molecular mechan-
ics) method of Gordon and coworkers (159), which is an extension of the IMOMM
method of Maseras & Morokuma (160) to deal with surfaces. Alternatively, there are
schemes in which the QM/MM boundary is chosen to pass through atoms, which are
split into pseudoatoms common to both the QM and MM regions. Recent examples
of this include a study of lattice deformations around a neutral oxygen vacancy in -
quartz (161) and an extension of the elastic polarizable environment model for ionic
solids (see above) to treat covalent solids (162). Effective model atomic potentials
have also been proposed to represent boundary atoms (166).
To date, comparatively fewer attempts to apply a QM/MM-type strategy for met-
als have appeared. The existing ones (163165) are all based on the density-based
embedding scheme discussed in Section 6.3. The background region is modeled with
classical EAM (embedded atom method) potentials. The background density
II
is
constructed as a superposition of atomic DFT densities computed for the free atom,
and Equation 32 is used to dene the coupling between the QM and MM regions.
7. CONCLUSIONS
As should now be evident, the palette of solid-state electronic structure techniques
is richly textured and substantive. We believe this is the sign of a healthy eld of
science, in which multiple strategies are simultaneously developed to probe the same
phenomena or states of matter. Even for formally exact theories such as DFT or
QMC, approximations are necessary to render them usable. Because all practical
implementations therefore involve some level of approximation, it is our strongly
held belief that no one technique should ever dominate to the exclusion of all others.
Competing techniques in theory, similar to competing experimental instruments,
provide multiple windows on phenomena that will always give a more complete
picture in the end.
Of late, DFT has been anointed as the method of choice for the quantum descrip-
tion of molecules and materials. Although often sufcient and even remarkably accu-
rate at times, it is best trusted only as a qualitative indicator because its quantitative
accuracy is still (and probably always will be) limited by approximate XC functionals.
Moreover, there are cases in which DFT fails completely, even in a qualitative sense.
Photochemistry, optical and photoemission spectra, HOMO-LUMOand band gaps,
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
physisorption (even chemisorption at times), strongly correlated materials, molecular
solids, biomolecules, and polymers are all at best poorly described within standard
implementations of DFT, even with the most accurate XCfunctionals available today
(GGA, meta-GGA, etc.).
For photochemistry and optical and photoemission spectra, the issue is that DFT
is a ground-state theory and these phenomena involve excited states. Therefore, one
should employ theories that properly account for electron excitations (embedded CI,
TDDFT, GW, or BSE). As outlined above, approximations inherent in TDDFT
(lack of XC functionals designed for excited states, problems applying it to periodic
systems), GW and Bethe-Salpeter (choice of self-energy evaluation, self-consistency,
and expense), and embedded CI (limited to local excitations, use of an approximate
kinetic energy density functional) still limit the impact of these methods somewhat.
Clearly, much room exists for improvements to all these methods in terms of algo-
rithmic efciency as well as breakthroughs to reduce the extent of approximations
made.
For HOMO-LUMO and band gaps, chemisorption such as CO on transition
metals, and strongly correlated electron materials, the failure of DFT results from
the approximate XC functional. Without exact exchange as in HF theory, a single
electron interacts with itself in DFT. In HF, this spurious self-Coulomb repulsion is
exactly canceled by an exchange energy. Without the exact nonlocal exchange term,
however, some self-repulsion remains. As a result, the LUMOs or conduction band
minima are articially too low, and some semiconductors become metals by accident
in an attempt to reduce electron repulsion by forming a more delocalized electron
distribution. Hybrid functionals that mix in some exact exchange counteract this
effect, as does the DFT + U and embedded CI methods. Again, it is good for the
eld to have multiple approaches to probe the same phenomena.
Polymers, biomolecules, molecular solids, and physisorption all are dominated by
dispersion forces, which again are not described by ground-state DFT. Dispersion in-
volves induced dipoleinduced dipole interactions, which can be thought of as result-
ing from two simultaneous single-electron excitations. Periodic Mller-Plesset per-
turbation theory is likely to be one of the methods of choice for such systems, as well as
embedded CI calculations, at least for the physisorption case. Progress in O(N) QMC
algorithms may also provide another excellent option for studying such systems.
As we look to the future, we are enthusiastic that these theories, as well as newones
to come, will continue to evolve in accuracy, efciency, and generality. In addition,
we envision that these approaches will also make connections to higher length scale
theories that hunger at the moment for information from the atomic scale. A truly
gestalt understanding of solid-state behavior will undoubtedly result.
SUMMARY POINTS
1. KS-DFT provides a good qualitative description of ground-state properties
for simple metals and covalently bonded materials. Interpretation of KS
quantities for excited states should be made with caution.
280 Huang
Carter
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
2. KS-DFTfails for strongly correlated materials (e.g., transition metal oxides)
due to the self-interaction error in approximate XC functionals. Methods
that incorporate exact exchange (hybrid functionals, LDA +U, and embed-
ded CI) can alleviate these problems.
3. Excited states are rigorously described within the many-body Greens
function formulation. The GW approximation can provide photoemis-
sion/inverse photoemission spectra and solution of the BSE yields opti-
cal spectra for insulators, semiconductors, and metals. Both methods yield
descriptions of excited states that are signicantly improved over a naive
interpretation of KS eigenvalue differences.
4. Correlation effects beyond DFT, as well as excited states, can be achieved
withQMC. However, QMCcalculations are demanding, and the availability
of a good trial function is critical. Most solid-state applications of QMChave
been restricted to simple, covalently bonded materials. To date, a fewresults
for transition metal oxides have been reported, and little to no results have
been reported for metallic crystals.
5. Embedding models are a highly versatile strategy for the description of
localized phenomena, and they allow for a local renement of theory in the
embedded region of interest. They can grow to be complex, however, and
special care must be taken to ensure that results are free of model bias.
DISCLOSURE STATEMENT
The authors are not aware of any biases that might be perceived as affecting the
objectivity of this review.
ACKNOWLEDGMENTS
We gratefully thank the Department of Energy, Basic Energy Sciences, and the Na-
tional Science Foundation for support.
LITERATURE CITED
1. Parr RG, Yang W. 1989. Density-Functional Theory of Atoms and Molecules. Int.
Ser. Monogr. Chem. New York: Oxford Univ. Press
2. Jones RO, Gunnarsson O. 1989. The density functional formalism, its applica-
tions and prospects. Rev. Mod. Phys. 61:689
3. Dreizler RM, Gross EKU. 1990. Density Functional Theory: An Approach to the
Quantum Many-Body Problem. Berlin: Springer-Verlag
4. AryasetiawanF, GunnarssonO. 1998. The GWmethod. Rep. Prog. Phys. 61:237
312
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
5. Hedin L. 1999. On correlation effects in electron spectroscopies and the GW
approximation. J. Phys. Condens. Matter 11:R489528
6. Onida G, Reining L, Rubio A. 2002. Electronic excitations: density-functional
vs many-body Greens-function approaches. Rev. Mod. Phys. 74:60159
7. Marques MAL, Ullrich CA, Nogueira F, Rubio A, Burke K, Gross EKU,
eds. 2006. Time-Dependent Density Functional Theory. Lect. Notes Phys. Berlin:
Springer
8. Hammond BL, Lester WA Jr, Reynolds PJ. 1994. Monte Carlo Methods in Ab
Initio Quantum Chemistry. Singapore: World Sci.
9. Foulkes WMC, Mitas L, Needs RJ, Rajagopal G. 2001. Quantum Monte Carlo
simulations of solids. Rev. Mod. Phys. 73:3383
10. Towler MD. 2006. The quantum Monte Carlo method. Phys. Status Solidi B
243:257398
11. Kotliar G, Savrasov SY, Haule K, Oudovenko VS, Parcollet O, Marianetti CA.
2006. Electronic structure calculations with dynamical mean-eld theory. Rev.
Mod. Phys. 78:865951
12. Sun JQ, Bartlett RJ. 1996. Second-order many-body perturbation-theory cal-
culations in extended systems. J. Chem. Phys. 104:855365
13. Ayala PY, Kudin KN, Scuseria GE. 2001. Atomic orbital Laplace-transformed
second-order Mller-Plesset theory for periodic systems. J. Chem. Phys.
115:9698707
14. Bezugly V, Birkenheuer U. 2004. Multireference conguration interaction
treatment of excited-state electron correlation in periodic systems: the band
structure of trans-polyacetylene. Chem. Phys. Lett. 399:5761
15. Pisani C, Busso M, Capecchi G, Casassa S, Dovesi R, et al. 2005. Local-
MP2 electron correlation method for nonconducting crystals. J. Chem. Phys.
122:094113
16. Goedecker S. 1999. Linear scaling electronic structure methods. Rev. Mod. Phys.
71:1085123
17. Hohenberg P, Kohn W. 1964. Inhomogeneous electron gas. Phys. Rev.
136:B864
18. Kohn W, Sham LJ. 1965. Self-consistent equations including exchange and
correlation effects. Phys. Rev. 140:A113338
19. Wang YA, Carter EA. 2000. Orbital-free kinetic energy density functional the-
ory. In Theoretical Methods in Condensed Phase Chemistry. Vol. 5: Progress in The-
oretical Chemistry and Physics, ed. SD Schwartz, p. 11784. Boston: Kluwer
20. Watson SC, Carter EA. 2000. Linear scaling parallel algorithms for the rst
principles treatment of metals. Comp. Phys. Commun. 128:6792
21. Perdew JP, Levy M. 1983. Physical content of the exact Kohn-Sham orbital
energies: band gaps and derivative discontinuities. Phys. Rev. Lett. 51:188487
22. Sham LJ, Schl uter M. 1983. Density-functional theory of the energy gap. Phys.
Rev. Lett. 51:188891
23. Sham LJ, Schl uter M. 1985. Density-functional theory of the band gap. Phys.
Rev. B 32:388389
24. Godby RW, Schl uter M, Sham LJ. 1988. Self-energy operators and exchange-
correlation potentials in semiconductors. Phys. Rev. B 37:1015975
282 Huang
Carter
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
25. Dirac PAM. 1930. Note on exchange phenomena in the Thomas atom. Proc.
Camb. Philos. Soc. 26:37685
26. Vosko SH, Wilk L, Nusair M. 1980. Accurate spin-dependent electron liquid
correlation energies for local spin density calculations: a critical analysis. Can.
J. Phys. 58:120011
27. Perdew JP, Zunger A. 1981. Self-interaction correction to density-functional
approximations for many-electron systems. Phys. Rev. B 23:504879
28. Perdew JP, Wang Y. 1992. Accurate and simple analytic representation of the
electron-gas correlation energy. Phys. Rev. B 45:1324449
29. Ceperley DM, Alder BJ. 1980. Ground state of the electron gas by a stochastic
method. Phys. Rev. Lett. 45:56669
30. Gunnarsson O, Lundqvist BI. 1976. Exchange and correlation in atoms,
molecules, and solids by the spin-density-functional formalism. Phys. Rev. B
13:427498
31. Gunnarsson O, Lundqvist BI. 1977. Erratum: exchange and correlation in
atoms, molecules, and solids by the spin-density-functional formalism. Phys.
Rev. B 15:6006
32. Gunnarsson O, Jonson M, Lundqvist BI. 1977. Exchange and correlation in
inhomogeneous electron systems. Solid State Commun. 24:76568
33. Burke K, Perdew JP, Ernzerhof M. 1998. Why semilocal functionals work:
accuracy of the on-top pair density and importance of system averaging.
J. Chem. Phys. 109:376071
34. Perdew JP. 1991. In Electronic Structure of Solids 91. ed. P Ziesche, H Eschrig,
p. 11. Berlin: Akad. Verlag
35. Perdew JP, Chevary JA, Vosko SH, Jackson KA, Pederson MR, et al. 1992.
Atoms, molecules, solids, and surfaces: applications of the generalized gradient
approximation for exchange and correlation. Phys. Rev. B 46:667187
36. Perdew JP, Chevary JA, Vosko SH, Jackson KA, Pederson MR, et al. 1993.
Erratum: atoms, molecules, solids, and surfaces: applications of the generalized
gradient approximation for exchange and correlation. Phys. Rev. B 48:4978
37. Perdew JP, Burke K, Ernzerhof M. 1996. Generalized gradient approximation
made simple. Phys. Rev. Lett. 77:386568
38. Perdew JP, Burke K, Ernzerhof M. 1997. Erratum: generalized gradient ap-
proximation made simple. Phys. Rev. Lett. 78:1396
39. Zhang Y, Yang W. 1998. Comment on Generalized gradient approximation
made simple. Phys. Rev. Lett. 80:890
40. Perdew JP, Burke K, Ernzerhof M. 1998. Perdew, Burke, and Ernzerhof reply.
Phys. Rev. Lett. 80:891
41. Hammer B, Hansen LB, Nrskov JK. 1999. Improved adsorption energetics
within density-functional theory using revised Perdew-Burke-Ernzerhof func-
tionals. Phys. Rev. B 59:741321
42. Kurth S, Perdew JP, Blaha P. 1999. Molecular and solid-state tests of den-
sity functional approximations: LSD, GGAs, and meta-GGAs. Int. J. Quantum
Chem. 75:889909
43. von Barth U. 2004. Basic density-functional theory: an overview. Phys. Scr.
T109:939
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
44. Staroverov VN, Scuseria GE, Tao J, PerdewJP. 2004. Tests of a ladder of density
functionals for bulk solids and surfaces. Phys. Rev. B 69:075102
45. Hamann DR. 1978. Semiconductor charge densities with hard-core and soft-
core pseudopotentials. Phys. Rev. Lett. 42:66265
46. Bachelet GB, Christensen NE. 1985. Relativistic and core-relaxation effects on
the energy bands of gallium arsenide and germanium. Phys. Rev. B 31:87987
47. Perdew JP, Kurth S, Zupan A, Blaha P. 1999. Accurate density functional with
correct formal properties: a step beyond the generalized gradient approxima-
tion. Phys. Rev. Lett. 82:254447
48. Perdew JP, Ernzerhof M, Burke K. 1996. Rationale for mixing exact exchange
with density functional approximations. J. Chem. Phys. 105:998285
49. Bylander DM, Kleinman L. 1990. Good semiconductor band gaps with a mod-
ied local-density approximation. Phys. Rev. B 41:786871
50. Seidl A, G orling A, Vogl P, Majewski JA, Levy M. 1996. Generalized Kohn-
Sham schemes and the band-gap problem. Phys. Rev. B 53:376474
51. Heyd J, Scuseria GE, Ernzerhof M. 2003. Hybrid functionals based on a
screened Coulomb potential. J. Chem. Phys. 118:820715
52. Heyd J, Scuseria GE, Ernzerhof M. 2006. Erratum: hybrid functionals based
on a screened Coulomb potential. J. Chem. Phys. 124:219906
53. Heyd J, Peralta JE, Scuseria GE. 2005. Energy band gaps and lattice param-
eters evaluated with the Heyd-Scuseria-Ernzerhof screened hybrid functional.
J. Chem. Phys. 123:174101
54. Paier J, Marsman M, Hummer K, Kresse G, Gerber IC,

Angy an JG. 2006.
Screened hybrid density functionals applied to solids. J. Chem. Phys. 124:154709
55. Paier J, Marsman M, Hummer K, Kresse G, Gerber IC,

Angy an JG. 2006.
Erratum: Screened hybrid density functionals applied to solids. J. Chem. Phys.
125:249901
56. Liechtenstein AI, Anisimov VI, Zaanen J. 1995. Density-functional theory and
strong interactions: orbital ordering in Mott-Hubbard insulators. Phys. Rev. B
52:R546770
57. Anisimov VI, Gunnarsson O. 1991. Density-functional calculation of effective
Coulomb interactions in metals. Phys. Rev. B 43:757074
58. Mosey NJ, Carter EA. 2007. Ab initio evaluation of Coulomb and exchange
parameters for DFT + U calculations. Phys. Rev. B 76:155123
59. Hedin L. 1965. New method for calculating the one-particle Greens function
with application to the electron-gas problem. Phys. Rev. 139:A796823
60. Hybertsen MS, Louie SG. 1985. Electron correlation in semiconductors and
insulators: band gaps and quasiparticle energies. Phys. Rev. B 34:5390413
61. Hybertsen MS, Louie SG. 1985. First-principles theory of quasiparticles: calcu-
lation of band gaps in semiconductors and insulators. Phys. Rev. Lett. 55:141821
62. White ID, Godby RW, Rieger MM, Needs RJ. 1998. Dynamic image potential
at an Al(111) surface. Phys. Rev. Lett. 80:426568
63. Rohlng M, Louie SG. 2000. Electron-hole excitations and optical spectra from
rst principles. Phys. Rev. B 62:492744
64. Rohlng M. 2000. Excited states of molecules from Greens function perturba-
tion techniques. Int. J. Quantum Chem. 80:80715
284 Huang
Carter
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
65. Pulci O, Reining L, Onida G, Solec RD, Bechstedt F. 2001. Many-body ef-
fects on one-electron energies and wave functions in low-dimensional systems.
Comput. Mater. Sci. 20:3004
66. Aryasetiawan F. 1992. Self-energy of ferromagnetic nickel in the GW approxi-
mation. Phys. Rev. B 46:1305164
67. Lundqvist BI. 1967. Single particle spectrum of degenerated electron gases. I.
Structure of spectral weight function. Phys. Kondens. Mater. 6:193205
68. Lundqvist BI. 1967. Single particle spectrum of degenerated electron gases. II.
Numerical results for electrons coupled with plasmons. Phys. Kondens. Mater.
6:20617
69. Steinbeck L, Rubio A, Reining L, Torrent M, White ID, Godby RW. 2000.
Enhancements to the GW space-time method. Comp. Phys. Commun. 125:105
18
70. Miyake T, Aryasetiawan F. 2000. Efcient algorithm for calculating noninter-
acting frequency-dependent linear response functions. Phys. Rev. B 61:717275
71. Leb` egue S, Arnaud B, Alouani M, Bloechl PE. 2003. Implementation of an all-
electron GW approximation based on the projector augmented wave method
without plasmon pole approximation: application to Si, SiC, AlAs, InAs, NaH,
and KH. Phys. Rev. B 67:155208
72. Shishkin M, Kresse G. 2006. Implementation and performance of the
frequency-dependent GW method within the PAW framework. Phys. Rev. B
74:035101
73. Patterson CH. 2006. Comparison of hybrid density functional, Hartree-Fock,
and GW calculations on NiO. Int. J. Quantum Chem. 106:338386
74. van Schilfgaarde M, Kotani T, Faleev SV. 2006. Adequacy of approximations
in GW theory. Phys. Rev. B 74:245125
75. Sch one W-D, Eguiluz AG. 1998. Self-consistent calculations of quasiparticle
states in metals and semiconductors. Phys. Rev. Lett. 81:166265
76. Tamme D, Schepe R, Henneberger K. 1999. Comment on Self-consistent
calculations of quasiparticle states in metals and semiconductors. Phys. Rev.
Lett. 83:241
77. Eguiluz AG. 1999. Eguiluz replies. Phys. Rev. Lett. 83:242
78. Ku W, Eguiluz AG. 2002. Band-gap problem in semiconductors revisited: ef-
fects of core states and many-body self-consistency. Phys. Rev. Lett. 89:126401
79. Delaney K, Garca-Gonz alez P, Rubio A, Rinke P, Godby RW. 2004. Comment
on Band-gap problem in semiconductors revisited: effects of core states and
many-body self-consistency. Phys. Rev. Lett. 93:249701
80. Ku W, Eguiluz AG. 2004. Ku and Eguiluz reply. Phys. Rev. Lett. 93:249702
81. Faleev SV, vanSchilfgaarde M, Kotani T. 2004. All-electronself-consistent GW
approximation: application to Si, MnO, and NiO. Phys. Rev. Lett. 93:126406
82. van Schilfgaarde M, Kotani T, Faleev SV. 2006. Quasiparticle self-consistent
GW theory. Phys. Rev. Lett. 96:226402
83. von Barth U, Holm B. 1996. Self-consistent GW
0
results for the electron gas:
xed screened potential W
0
within the random-phase approximation. Phys. Rev.
B 54:841119
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
84. von Barth U, Holm B. 1997. Erratum: self-consistent GW
0
results for the elec-
tron gas: xed screened potential W
0
within the random-phase approximation.
Phys. Rev. B 55:1012022
85. HolmB, vonBarthU. 1998. Fully self-consistent GWself-energy of the electron
gas. Phys. Rev. B 57:210817
86. Holm B. 1999. Total energies from GW calculations. Phys. Rev. Lett. 83:78891
87. Hamada N, Hwang M, Freeman AJ. 1990. Self-energy correction for the energy
bands of silicon by the full-potential linearized augmented-plane-wave method:
effect of the valence-band polarization. Phys. Rev. B 41:362026
88. Kotani T, van Schilfgaard M. 2002. All-electron GW approximation with the
mixed basis expansion based on the full-potential LMTO method. Solid State
Commun. 121:46165
89. Chelikowsky JR, Kronik L, Vasiliev I. 2003. Time-dependent density-functional
calculations for the optical spectra of molecules, clusters, and nanocrystals.
J. Phys. Condens. Matter 15:R151747
90. Burke K, Werschnik J, Gross EKU. 2005. Time-dependent density functional
theory: past, present, and future. J. Chem. Phys. 123:062206
91. Sottile F, Bruneval F, Marinopoulos AG, Dash LK, Botti S, et al. 2005. TDDFT
from molecules to solids: the role of long-range interactions. Int. J. Quantum
Chem. 102:684701
92. Elliott P, Burke K, Furche F. 2007. Excited states from time-dependent density
functional theory. Manuscript under review. arXiv:cond-mat/0703590
93. Botti S, Schindlmayr A, DelSole R, Reining L. 2007. Time-dependent density-
functional theory for extended systems. Rep. Prog. Phys. 70:357407
94. Casida ME. 1996. Time-dependent density functional response theory of
molecular systems: theory, computational methods, and functionals. In Recent
Developments and Applications of Modern Density Functional Theory, Vol. 4: Theo-
retical and Computational Chemistry, ed. JM Seminario, p. 391440. Amsterdam:
Elsevier Sci.
95. Jamorski C, Casida ME, Salahub DR. 1996. Dynamic polarizabilities and ex-
citation spectra from a molecular implementation of time-dependent density-
functional response theory: N
2
as a case study. J. Chem. Phys. 104:513447
96. Onida G, Reining L, Godby RW, Sole RD, Andreoni W. 1995. Ab initio cal-
culations of the quasiparticle and absorption spectra of clusters: the sodium
tetramer. Phys. Rev. Lett. 75:81821
97. Albrecht S, Onida G, Reining L. 1997. Ab initio calculation of the quasiparticle
spectrum and excitonic effects in Li
2
O. Phys. Rev. B 55:1027881
98. Vignale G, Kohn W. 1996. Current-dependent exchange-correlation potential
for dynamical linear response theory. Phys. Rev. Lett. 77:203740
99. Berger JA, de Boeij PL, van Leeuwen R. 2007. Analysis of the Vignale-Kohn
current functional in the calculation of the optical spectra of semiconductors.
Phys. Rev. B 75:035116
100. Kalos MH, Whitlock PA. 1986. Monte Carlo Methods, Vol. 1. New York: Wiley
& Sons
101. Williamson AJ, Hood RQ, Grossman JC. 2001. Linear-scaling quantumMonte
Carlo calculations. Phys. Rev. Lett. 87:246406
286 Huang
Carter
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
102. Alf` e D, Gillan MJ. 2004. Linear-scaling quantum Monte Carlo technique with
nonorthogonal localized orbitals. J. Phys. Condens. Matter 16:L30511
103. Grossman JC, Mitas L, Raghavachari K. 1995. Structure and stability of molec-
ular carbon: importance of electron correlation. Phys. Rev. Lett. 75:387073
104. Grossman JC, Mitas L, Raghavachari K. 1996. Erratum: structure and stability
of molecular carbon: importance of electron correlation. Phys. Rev. Lett. 76:1006
105. Mitas L. 1996. Electronic structure by quantumMonte Carlo: atoms, molecules
and solids. Comp. Phys. Commun. 96:10717
106. Williamson AJ, Hood RQ, Needs RJ, Rajagopal G. 1998. Diffusion quantum
Monte Carlo calculations for excited states of silicon. Phys. Rev. B 57:1214044
107. Williamson AJ, Grossman JC, Hood RQ, Puzder A, Galli G. 2002. Quantum
Monte Carlo calculations of nanostructure optical gaps: application to silicon
quantum dots. Phys. Rev. Lett. 89:196803
108. Drummond ND, Williamson AJ, Needs RJ, Galli G. 2005. Electron emission
from diamondoids: a diffusion quantum Monte Carlo study. Phys. Rev. Lett.
95:096801
109. Grossman JC. 2003. Benchmark quantum Monte Carlo calculations. J. Chem.
Phys. 117:143440
110. Burkatzki M, Filippi C, Dolg M. 2007. Energy-consistent pseudopotentials for
quantum Monte Carlo calculations. J. Chem. Phys. 126:234105
111. Tanaka S. 1995. Variational quantum Monte-Carlo approach to the electronic
structure of NiO. J. Phys. Soc. Jpn. 64:427077
112. Needs RJ, Towler MD. 2003. The diffusion quantum Monte Carlo method:
designing trial wave functions for NiO. Int. J. Mod. Phys. B 17:542534
113. Mit a s L, Shirley EL, Ceperley DM. 1991. Nonlocal pseudopotentials and dif-
fusion Monte Carlo. J. Chem. Phys. 95:346775
114. Filippi C, Umrigar CJ. 2000. Correlated sampling in quantum Monte Carlo: a
route to forces. Phys. Rev. B 61:R1629194
115. Casalegno M, Mella M, Rappe AM. 2003. Computing accurate forces in quan-
tum Monte Carlo using Pulays corrections and energy minimization. J. Chem.
Phys. 118:7193201
116. Sauer J. 1989. Molecular models in ab initio studies of solids and surfaces: from
ionic crystals and semiconductors to catalysts. Chem. Rev. 89:199255
117. Pacchioni G, Bagus PS, Parmigiani F, eds. 1992. Cluster Models for Surface and
Bulk Phenomena. NATO ASI Ser. New York: Plenum
118. Whitten JL, Yang H. 1996. Theory of chemisorption and reactions on metal
surfaces. Surf. Sci. Rep. 24:55124
119. Jug K, Bredow T. 2004. Models for the treatment of crystalline solids and
surfaces. J. Comput. Chem. 25:155167
120. Evjen HM. 1932. On the stability of certain heteropolar crystals. Phys. Rev.
39:67587
121. Sousa C, Casanovas J, Rubio J, Illas F. 1993. Madelung elds from optimized
point charges for abinitiocluster model calculations onionic systems. J. Comput.
Chem. 14:68084
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
122. Dungsrikaew V, Limtrakul J, Hermansson K, Probst M. 2004. Comparison of
methods for point-charge representation of electrostatic elds. Int. J. Quantum
Chem. 96:1722
123. Winter NW, Pitzer RM, Temple DK. 1987. Theoretical study of a Cu
+
ion
impurity in a NaF host. J. Chem. Phys. 86:354956
124. Barandiar anZ, SeijoL. 1988. The abinitiomodel potential representationof the
crystalline environment: theoretical study of the local distortion on NaCl:Cu
+
.
J. Chem. Phys. 89:573946
125. Danyliv O, Kantorovich L. 2004. Comparison of localization procedures for
applications in crystal embedding. Phys. Rev. B 70:075113
126. Pisani C. 1978. Approach to the embedding problem in chemisorption in a
self-consistent-eld-molecular-orbital formalism. Phys. Rev. B 17:314353
127. Pisani C, Dovesi R, Carosso P. 1979. Moderately-large-embedded-cluster ap-
proach to the study of local defects in solids: vacancy and substitutional impu-
rities in graphite. Phys. Rev. B 20:534557
128. Pisani C, Dovesi R, Nada R, Kantorovich LN. 1990. Ab initio Hartree-Fock
perturbed-cluster treatment of local defects in crystals. J. Chem. Phys. 92:7448
60
129. Kr uger S, R osch N. 1994. The moderately-large-embedded-cluster method for
metal surfaces: a density-functional study of atomic adsorption. J. Phys. Condens.
Matter 6:814966
130. Pisani C, Dovesi R, Roetti C, Caus a M, Orlando R, et al. 2000. CRYSTAL and
EMBED, two computational tools for the ab initio study of electronic properties
of crystals. Int. J. Quantum Chem. 77:103248
131. Ingleseld JE. 2001. Embedding at surfaces. Comp. Phys. Commun. 137:89107
132. Bormet J, Neugebauer J, Schefer M. 1994. Chemical trends and bonding
mechanisms for isloated adsorbates on Al(111). Phys. Rev. B 49:1724252
133. Whitten JL, Yang H. 1995. Theoretical studies of surface reactions on metals.
Int. J. Quantum Chem. 56:4147
134. GutdeutschU, Birkenheuer U, Kr uger S, R oschN. 1997. Oncluster embedding
schemes based on orbital space partitioning. J. Chem. Phys. 106:602030
135. Cortona P. 1991. Self-consistently determined properties of solids without
band-structure calculations. Phys. Rev. B 44:845458
136. Wesolowski TA, Warshel A. 1993. Frozen density functional approach for ab
initio calculations of solvated molecules. J. Phys. Chem. 97:805053
137. Neugebauer J, Louwerse MJ, Baerends EJ. 2005. The merits of the frozen-
density embedding scheme to model solvatochromic shifts. J. Chem. Phys.
122:094115
138. Wesolowski TA, Ellinger Y, Weber J. 1998. Density functional theory with an
approximate kinetic energy functional applied to study structure and stability
of weak van der Waals complexes. J. Chem. Phys. 108:607883
139. Stefanovich EV, Truong TN. 1996. Embedded density functional approach for
calculations of adsorption on ionic crystals. J. Chem. Phys. 104:294655
140. Trail JR, Bird DM. 2000. Density-functional embedding using a plane-wave
basis. Phys. Rev. B 62:1640211
288 Huang
Carter
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
141. Govind N, Wang YA, da Silva AJR, Carter EA. 1998. Accurate ab initio energet-
ics of extended systems via explicit correlation embedded in a density functional
environment. Chem. Phys. Lett. 295:12934
142. Govind N, Wang YA, Carter EA. 1999. Electronic-structure calculations
by rst-principles density-based embedding of explicitly correlated systems.
J. Chem. Phys. 110:767788
143. Kl uner T, Govind N, Wang YA, Carter EA. 2001. Prediction of electronic
excited states of adsorbates on metal surfaces from rst principles. Phys. Rev.
Lett. 86:595457
144. Kl uner T, Govind N, Wang YA, Carter EA. 2002. Reply to comment on pre-
diction of electronic excited states of adsorbates on metal surfaces from rst
principles. Phys. Rev. Lett. 88:209702
145. Kl uner T, Govind N, Wang YA, Carter EA. 2002. Periodic density functional
embedding theory for complete active space self-consistent eld and congura-
tioninteractioncalculations: groundandexcitedstates. J. Chem. Phys. 116:4254
146. Huang P, Carter EA. 2006. Local electronic structure around a single Kondo
impurity. Nano Lett. 6:114650
147. Huang P, Carter EA. 2006. Self-consistent embedding theory for locally cor-
related conguration interaction wave functions in condensed matter. J. Chem.
Phys. 125:084102
148. Huang P, Carter EA. 2007. Ab initio explanation of tunneling lineshapes for
the Kondo impurity state. Submitted
149. Sharifzadeh S, Huang P, Carter EA. 2007. Embedded cluster description of CO
on Cu(111): resolution of the site preference conundrum. Submitted
150. Lahav D, Kl uner T. 2007. A self-consistent density based embedding scheme
applied to the adsorption of CO on Pd(111). J. Phys. Condens. Matter 19:226001
151. Lin H, Truhlar DG. 2007. QM/MM: What have we learned, where are we, and
where do we go from here? Theor. Chem. Acc. 117:18599
152. Vail JM. 1990. Theory of electronic defects: applications to MgO and alkali
halides. J. Phys. Chem. Solids 51:589607
153. Dick BG Jr, Overhauser AW. 1958. Theory of the dielectric constants of alkali
halide crystals. Phys. Rev. 112:90103
154. Pascual JL, Seijo L. 1995. Ab initio model potential embedded cluster calcula-
tions including lattice relaxation and polarization: local distortions on Mn
2+
-
doped CaF
2
. J. Chem. Phys. 102:536876
155. Nasluzov VA, Rivanenkov VV, Gordienko AB, Neyman KM, Birkenheuer U,
R osch N. 2001. Cluster embedding in an elastic polarizable environment: den-
sity functional study of Pd atoms adsorbed at oxygen vacancies of MgO(001).
J. Chem. Phys. 115:815771
156. Redondo A, Goddard WA III, McGill TC, Surratt GT. 1976. Relaxation of
(111) silicon surface atoms from studies of Si
4
H
9
clusters. Solid State Commun.
20:73336
157. Carmer CS, Weiner B, Frenklach M. 1993. Molecular dynamics with combined
quantum and empirical potentials: C
2
H
2
adsorption on Si(100). J. Chem. Phys.
99:135672
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
158. Sauer J, Sierka M. 2000. Combining quantum mechanics and interatomic po-
tential functions in ab initio studies of extended systems. J. Comput. Chem.
21:147093
159. Shoemaker JR, Burggraf LW, Gordon MS. 1999. SIMOMM: an integrated
molecular orbital/molecular mechanics optimization scheme for surfaces.
J. Phys. Chem. A 103:324551
160. Maseras F, Morokuma K. 1995. IMOMM: a new integrated ab initio + molec-
ular mechanics geometry optimization scheme of equilibrium structures and
transition states. J. Comput. Chem. 16:117079
161. Sulimov VB, Sushko PV, Edwards AH, Shluger AL, Stoneham AM. 2002.
Asymmetry and long-range character of lattice deformation by neutral oxygen
vacancy in -quartz. Phys. Rev. B 66:024108
162. Nasluzov VA, Ivanova EA, Shor AM, Vayssilov GN, Birkenheuer U, R osch N.
2003. Elastic polarizable environment cluster embedding approach for covalent
oxides and zeolites based on a density functional method. J. Phys. Chem. B
107:222841
163. Choly N, Lu GEW, Kaxiras E. 2005. Multiscale simulations in simple metals:
a density-functional-based methodology. Phys. Rev. B 71:094101
164. Lu G, Tadmor EB, Kaxiras E. 2006. From electrons to nite elements: a con-
current multiscale approach for metals. Phys. Rev. B 73:024108
165. Liu Y, Lu G, Chen Z, Kioussis N. 2007. An improved QM/MM approach for
metals. Model. Simul. Mater. Sci. Eng. 15:27584
166. Slav cek S, Martnez TJ. 2006. Multicentered valence electron effective poten-
tials: a solution to the link atom problem for ground and excited electronic
states. J. Chem. Phys. 124:084107
290 Huang
Carter
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
AR340-FM ARI 26 February 2008 21:26
Annual Review of
Physical Chemistry
Volume 59, 2008
Contents
A Fortunate Life in Physical Chemistry
Stuart A. Rice p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 1
Chemistry and Photochemistry of Mineral Dust Aerosol
David M. Cwiertny, Mark A. Young, and Vicki H. Grassian p p p p p p p p p p p p p p p p p p p p p p p p p p 27
Femtobiology
Villy Sundstr om p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 53
Structures, Kinetics, Thermodynamics, and Biological Functions
of RNA Hairpins
Philip C. Bevilacqua and Joshua M. Blose p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 79
Understanding Protein Evolution: From Protein Physics
to Darwinian Selection
Konstantin B. Zeldovich and Eugene I. Shakhnovich p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 105
Quasicrystal Surfaces
Patricia A. Thiel p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 129
Molecular Ordering and Phase Behavior of Surfactants at Water-Oil
Interfaces as Probed by X-Ray Surface Scattering
Mark L. Schlossman and Aleksey M. Tikhonov p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 153
Extraordinary Transmission of Metal Films with Arrays
of Subwavelength Holes
James V. Coe, Joseph M. Heer, Shannon Teeters-Kennedy, Hong Tian,
and Kenneth R. Rodriguez p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 179
The Ultrafast Dynamics of Photodetachment
Xiyi Chen and Stephen E. Bradforth p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 203
Energy Flow in Proteins
David M. Leitner p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 233
Advances in Correlated Electronic Structure Methods for Solids,
Surfaces, and Nanostructures
Patrick Huang and Emily A. Carter p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 261
Two-Dimensional Infrared Spectroscopy of Photoswitchable Peptides
Peter Hamm, Jan Helbing, and Jens Bredenbeck p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 291
ix
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
Wave-Packet Interferometry and Molecular State Reconstruction:
Spectroscopic Adventures on the Left-Hand Side of the
Schr odinger Equation
Jeffrey A. Cina p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 319
Ions at Aqueous Interfaces: From Water Surface to Hydrated Proteins
Pavel Jungwirth and Bernd Winter p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 343
Nanografting for Surface Physical Chemistry
Maozi Liu, Nabil A. Amro, and Gang-yu Liu p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 367
Extending X-Ray Crystallography to Allow the Imaging of
Noncrystalline Materials, Cells, and Single Protein Complexes
Jianwei Miao, Tetsuya Ishikawa, Qun Shen, and Thomas Earnest p p p p p p p p p p p p p p p p p p 387
Patterning Fluid and Elastomeric Surfaces Using Short-Wavelength
UV Radiation and Photogenerated Reactive Oxygen Species
Babak Sanii and Atul N. Parikh p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 411
Equation-of-Motion Coupled-Cluster Methods for Open-Shell
and Electronically Excited Species: The Hitchhikers Guide
to Fock Space
Anna I. Krylov p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 433
Attosecond Electron Dynamics
Matthias F. Kling and Marc J.J. Vrakking p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 463
Functional Polymer Brushes in Aqueous Media from Self-Assembled
and Surface-Initiated Polymers
Ryan Toomey and Matthew Tirrell p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 493
Electronic Spectroscopy of Carbon Chains
Evan B. Jochnowitz and John P. Maier p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 519
Multiscale Simulation of Soft Matter: From Scale Bridging
to Adaptive Resolution
Matej Praprotnik, Luigi Delle Site, and Kurt Kremer p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 545
Free Energies of Chemical Reactions in Solution and in Enzymes
with Ab Initio Quantum Mechanics/Molecular Mechanics Methods
Hao Hu and Weitao Yang p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 573
Fluctuation Theorems
E.M. Sevick, R. Prabhakar, Stephen R. Williams, and Debra J. Searles p p p p p p p p p p p p 603
Structure, Dynamics, and Assembly of Filamentous Bacteriophages
by Nuclear Magnetic Resonance Spectroscopy
Stanley J. Opella, Ana Carolina Zeri, and Sang Ho Park p p p p p p p p p p p p p p p p p p p p p p p p p p p p 635
Inside a Collapsing Bubble: Sonoluminescence and the Conditions
During Cavitation
Kenneth S. Suslick and David J. Flannigan p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 659
x Contents
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.
Elastic Modeling of Biomembranes and Lipid Bilayers
Frank L.H. Brown p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 685
Water in Nonpolar Connement: From Nanotubes to Proteins
and Beyond
Jayendran C. Rasaiah, Shekhar Garde, and Gerhard Hummer p p p p p p p p p p p p p p p p p p p p p 713
High-Resolution Spectroscopic Studies and Theory of Parity Violation
in Chiral Molecules
Martin Quack, Jrgen Stohner, and Martin Willeke p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 741
Collapse Mechanisms of Langmuir Monolayers
Ka Yee C. Lee p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 771
Indexes
Cumulative Index of Contributing Authors, Volumes 5559 p p p p p p p p p p p p p p p p p p p p p p p p 793
Cumulative Index of Chapter Titles, Volumes 5559 p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p p 796
Errata
An online log of corrections to Annual Review of Physical Chemistry articles may be
found at http://physchem.annualreviews.org/errata.shtml
Contents xi
A
n
n
u
.

R
e
v
.

P
h
y
s
.

C
h
e
m
.

2
0
0
8
.
5
9
:
2
6
1
-
2
9
0
.

D
o
w
n
l
o
a
d
e
d

f
r
o
m

w
w
w
.
a
n
n
u
a
l
r
e
v
i
e
w
s
.
o
r
g
b
y

U
n
i
v
e
r
s
i
d
a
d

A
u
t
o
n
o
m
a

d
e
l

E
s
t
a
d
o

d
e

M
o
r
e
l
o
s

o
n

1
1
/
3
0
/
1
2
.

F
o
r

p
e
r
s
o
n
a
l

u
s
e

o
n
l
y
.

Annual Reviews Solid

Caricato da

Informazioni sul documento

Copyright

Formati disponibili

Condividi questo documento

Condividi o incorpora il documento

Opzioni di condivisione

Hai trovato utile questo documento?

Questo contenuto è inappropriato?

Copyright:

Formati disponibili

Annual Reviews Solid

Caricato da

Copyright:

Formati disponibili

ANRV340-PC59-11 ARI 26 February 2008 21:0

and Emily A. Carter

Current afliation: Lawrence Livermore

, and the local kinetic energy density,

| appearing in the exchange operator. In the Thomas-Fermi theory of

; ). Physically, W represents the potential

, taking into account the polarization of the surround-

|, and thus we expect an expansion in

|. The leading term in the expansion

is replaced by the TDDFT XC kernel

above reduces to a one-electron integral over r. The widespread use

Potrebbero piacerti anche